Nonlinear Filter Modeling Diffusion in Target Heading/Speed

To improve automated target tracking and correlation for strategic defense, this proposal models the dynamic behavior of manuevering vehicles by allowing target heading and speed, or their derivatives, to undergo independent diffusion processes. Current methods for surveillance sensors model diffusion in geographically fixed coordinate systems and fail to exploit course/speed nonlinearities, and known dynamic constraints, over time. A stochastic PDE characterizes the transition density of the process, and is supplanted by a sequence of discrete stochastic difference problems. The discrete problems are solved numerically, and an optimal, nonlinear, point-mass filter combines transition density updates with nongaussian observation updates. Operations count, throughput analysis, and comparison with a related filter indicate the filter's feasibility. Applications include multisensor data fusion, runway incursion prevention, tracking in nonradar sectors, and ship collision avoidance.